library('ggplot2') # graphics library
library('tibble') # nice dataframesggplot
Note: This file is greatly based on the material prepared by Julian Langer and Lachlan Deer for the programming course for previous years
Making plots with ggplot2
I personally tend to forget the details about making graphs in R every time I have to make them, so here is a cheatsheet which may come in handy for when you have to make graphs, taken from the tidyverse package webpage: https://github.com/rstudio/cheatsheets/blob/main/data-visualization-2.1.pdf
The dataset
Lets load a dataset which is included in the ggplot library.
diamonds_df = as_tibble(diamonds)
head(diamonds_df) # look at the head of the dataset# A tibble: 6 × 10
carat cut color clarity depth table price x y z
<dbl> <ord> <ord> <ord> <dbl> <dbl> <int> <dbl> <dbl> <dbl>
1 0.23 Ideal E SI2 61.5 55 326 3.95 3.98 2.43
2 0.21 Premium E SI1 59.8 61 326 3.89 3.84 2.31
3 0.23 Good E VS1 56.9 65 327 4.05 4.07 2.31
4 0.29 Premium I VS2 62.4 58 334 4.2 4.23 2.63
5 0.31 Good J SI2 63.3 58 335 4.34 4.35 2.75
6 0.24 Very Good J VVS2 62.8 57 336 3.94 3.96 2.48Let’s briefly look at this dataset. It includes the prices and features of around 54 000 round cut diamonds. It contains the following variables (Hint: you can also look up the description of the dataset using ?diamonds once you loaded ggplot2):
- price: price in US dollars
- carat: weight of the diamond
- cut: quality of the cut (Fair, Good, Very Good, Premium, Ideal)
- color: diamond colour from J (worst) to D (best)
- clarity: clarity of the diamond (l1, SI1, SI2, VS1, VS2, VVS1, VVS2, IF)
- x: length in mm
- y: width in mm
- z: depth in mm
- depth: total depth percentage
- table: width of top of diamond relative to widest point
Basic example: an ordinary scatterplot
We will start to build up the logic of the ggplot2 command step by step. No matter which graphics you want to draw, you have to start as follows:
ggplot(data = diamonds_df)
As you can see, if you execute this command, this opens an empty graphics windows and nothing else happens. This make sense. So far we have only told ggplot which data we want to use for the plot.
To actually see something we have to add a layer. Say we want to produce a scatter plot that shows us the association between the variables carat and price.
ggplot(data = diamonds_df) +
geom_point(aes(x = carat , y = price))
What have we done here? We have added a new layer to our empty plot. The function geom_point specifies that we want to add a scatterplot layer. the aes function which is a shorthand for aesthetics. Aesthetic mappings describe how variables in the data are mapped to visual properties (aesthetics) of geoms.
In our case, we have specified to map the carat of the diamond to the x-axis and the price of the diamond to the y-axis. There seems to be a positive relationship between carat and price although it is noisier for higher carat levels.
Now let’s do a simple histogram of the prices of the diamonds
ggplot(data = diamonds_df) +
geom_histogram(aes(x = price))`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Aesthetics
Aesthetic mappings
Of course, we can do more by adding further variables to our plot. Say we want to color the dots according to the clarity category. For this, we have to add the color argument to the aesfunction.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price, color = clarity))
There are also other aesthetics and we could have mapped the variable into each of them. We could have also just manually set a value for the aesthetics/ Some aesthetics are:
- size of dots: either the name of a variable, or size in mm
- for
shape: either the name of a variable, or shape of a point as a number - for
alpha: either the name of a variable, or value between 0 and 1
Now we can use the fact that clarity is an ordered factor and color de dots differenty depending in whether clarity is above or below a certain threshold ‘VS2’:
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price, color = clarity > 'VS2'))
Facets: Subplots for different variable values
What if we want do not map all variables into aesthetics of one graph but want to have one graph for each value of a variable? For our dataset, we could for example be interested in how carat is associated with price for different values of the variable cut. For this, we need to split our plot into different facets (i.e. subplots) using the facet_wrap function.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price)) +
facet_wrap(. ~ cut, nrow = 2)
As you can see, for each value of the cut variable, we get a separate subplot. Note that you should pass a discrete variable to the facet_wrap function. The nrow argument specified into how many rows the plots are organized. We can also cross-tabulate subplots for different variables with the facet_grid function.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price)) +
facet_grid(vars(cut), vars(clarity))
# Divide into rows
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price)) +
facet_grid(clarity ~ .)
# Divide into columns
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price)) +
facet_grid( . ~ clarity)
Disgression on formulas
You may have noticed the expression “clarity ~ cut” in the previous line. What is this expression exactly? It’s a formula in R. These are mainly used in graphs (as above) and regressions (which we will see later).
A formula is of the form ‘y ~ x + z + k’. Y is the dependent variable while x,z,k are the independent variables.
Geoms
What kind of plot do you have in mind? The use of geoms.
So far we have just produced a scatterplot. We can of course also create other kinds of plots using different geometrical objects or geoms. For example, we could fit a smooth line to the data with geom_smooth.
ggplot(data = diamonds_df) +
geom_smooth(mapping = aes(x = carat, y = price))`geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'
Depending on the geometrical object, different aesthetics can be used. To see which aesthetics are available type ?geom_smooth. Moreover, you can also see which additional arguments are available and which default values are set. In the following, we will change the confidence intervals and map the variable cut to the color aesthetic.
ggplot(data = diamonds_df) +
geom_smooth(mapping = aes(x = carat, y = price, color = cut), level = 0.9)`geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'
Multiple geoms in one plot
We can also put multiple geoms in one graph by creating multiple layers.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price), size = 0.2, alpha = 0.5) +
geom_smooth(mapping = aes(x = carat, y = price))`geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'
Statistical tranformations with stats
When you use ggplot, there is always a statistical tranformation - a stat - of the data running in the background. Consider the following dataframe:
test_df <- tibble(
colors = c("green", "blue", "red", "black", "white"),
numbers = c(20, 40, 50, 66, 8)
)
head(test_df)# A tibble: 5 × 2
colors numbers
<chr> <dbl>
1 green 20
2 blue 40
3 red 50
4 black 66
5 white 8Imagine we want to create a bar plot with the hight of the bars being the value of “numbers”
ggplot(data = test_df) +
geom_bar(mapping = aes(x = colors))
Well, this is not what we wanted. The x axis plots the categories for ‘colors’. The variable count is plotted on the y-axis. But you will not find the variable count in our dataset. Ggplot created a new dataframe with the variable colors and the variable count for each category and plotted this to the figure. This is an example of the count stat!
However, we want the geom to use the data in the numbers variable for the y-axis. We need to use a different stat.
ggplot(data = test_df) +
geom_bar(mapping = aes(x = colors, y = numbers), stat = "identity")
We can also plot the mean by color. Let’s add two observations to the dataframe
test_df <- tibble(
colors = c("green", "blue", "red", "black", "white", "green", "green"),
numbers = c(20, 40, 50, 66, 8, 10, 10)
)ggplot(data = test_df) +
geom_bar(mapping = aes(x = colors, y = numbers), stat = "summary", fun = "mean")
Again, there are tons of statistical transformations in ggplot.
Positions
Ok, now it is time to talk about position adjustments. What happens in our histograms when we map a variable to the fill aesthetic?
ggplot(data = diamonds_df) +
geom_bar(mapping = aes(x = cut, fill = clarity), position = 'stack')
This is cool, right? The bars for each clarity category are stacked upon each other for each cut category. This is because ggplot automatically uses the stack position adjustement when drawing geom_bar. There are more options however.
If you instead use the fill position, bars are stacked upon each other, but each stack of bars is forced to have the same height.
ggplot(data = diamonds_df) +
geom_bar(mapping = aes(x = cut, fill = clarity), position = "fill")
Third, you can also use the dodge position to make a useful change to the bar plot. This places the bars next to each other.
ggplot(data = diamonds_df) +
geom_bar(mapping = aes(x = cut, fill = clarity), position = "dodge")
Coordinate systems
The default coordinate system used by ggplot is the Cartesian one (coord_cartesian). I will only briefly introduce some other useful commands here. Rest assured that there is again much more to say here.
You can quickly flip axes using the coord_flip command. Let’s just do this to one of our bar charts.
ggplot(data = diamonds_df) +
geom_bar(mapping = aes(x = cut)) +
coord_flip()
With the coord_trans function you plot data to a Cartesian coordinate system, where axes are transformed by a function.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price)) +
coord_trans(x = "log", y = "log")
Sometimes you want to have a fixed aspect ratio for your coordinate system, you can use the coord_fixed command to create a Cartesian coordinate system with a fixed aspect ratio between x and y units.
Labeling
So far, we looked at the construction of graphs but did not change labels. We can do this using the labs function. In this case, we add a title and labels for the x-axis and y-axis as well as the legend.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price, color = clarity)) +
labs(title = "Relationship between weight and price of diamonds",
x = "Weight in carat",
y = "Price in dollars",
color = "Clarity")
Scales
Scales control the details of how data values are translated to visual properties. Override the default scales to tweak details like the axis labels or legend keys, or to use a completely different translation from data to aesthetic. labs() and lims() are convenient helpers for the most common adjustments to the labels and limits.
We will talk here specifically about three things that we can tweak using scale functions: axis ticks, legend labels and color schemes.
To change the axis ticks, you have to pass the breaks argument with the desired breaks to the scales.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price, color = clarity)) +
labs(title = "Relationship between weight and price of diamonds",
x = "Weight in carat",
y = "Price in dollars",
color = "Clarity") +
scale_x_continuous(breaks = seq(0, 5, 0.5)) +
scale_y_continuous() +
scale_color_discrete()
You can also change the labels by passing the labels argument to the scale with the vector of desired labels.
Next, we want to change the labels for our legend. These are related to the color aesthetic. We can just pass the labels argument with the vector of desired labels. In this example, I just pass a vector of numbers.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price, color = clarity), size = 0.2) +
labs(title = "Relationship between weight and price of diamonds",
x = "Weight in carat",
y = "Price in dollars",
color = "Clarity") +
scale_x_continuous() +
scale_y_continuous() +
scale_color_discrete(labels = seq(1, 8))
Finally, what if you want to change the color of our graph? It is obviously related to the color aesthetic. In this case, we don’t have to pass an argument to the scale function but instead have to completely replace the scale.
# brewer scales at http://colorbrewer2.org
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price, color = clarity), size = 0.2) +
labs(title = "Relationship between weight and price of diamonds",
x = "Weight in carat",
y = "Price in dollars",
color = "Clarity") +
scale_x_continuous() +
scale_y_continuous() +
scale_color_brewer(palette = "Reds")
If you want to manually set colors, you can do so with the scale_color_manual function.
Themes
You can also change the complete non-data elements of your graph by applying the theme function. You have the following themes available:
theme_bw: white background with grid linestheme_classic: classic theme; axes but not grid linestheme_dark: dark background for contrasttheme_gray: default theme with grey backgroundtheme_light: light axes and grid linestheme_linedraw: only black linestheme_minimal: minimal theme, no backgroundtheme_void: empty theme, only geoms are visible
There is also a package called ggthemes which gives you a ton of other templates. In the following, we just switch our graph to the classic theme.
ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price, color = clarity), size = 0.2) +
labs(title = "Relationship between weight and price of diamonds",
x = "Weight in carat",
y = "Price in dollars",
color = "Clarity") +
scale_color_brewer(palette = "Reds") +
theme_classic()
Saving graphs
We can save the graphs using the ggsave function. Look up the help file for ggsave to see into which formats you can export your plot. Here is an example:
my_plot <- ggplot(data = diamonds_df) +
geom_point(mapping = aes(x = carat, y = price) )
ggsave(filename = "myplot.pdf", plot = my_plot)Saving 7 x 5 in image